AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

BD-Merging: Bias-Aware Dynamic Model Merging with Evidence-Guided Contrastive Learning

Yuhan Xie, Chen Lyu

CVPR 2026 · Label-efficient and robust learning › Recognition and long-tail · approach: Data augmentation and generalization bias · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Add debiased router (per-sample merging weights) (ablation)BD-Merging · 8 tasks, corrupted L2avg accuracy: 67.25 -> 78.78Improved
Add debiased router (per-sample merging weights) (ablation)BD-Merging · 8 tasks, cleanavg accuracy: 78.31 -> 87.15Improved
Add discrepancy-aware contrastive loss L_Dis (ablation)BD-Merging · 8 tasks, corrupted L2avg accuracy: 74.26 -> 78.78Improved
Add Adjacency Discrepancy Score (ADS) (ablation)BD-Merging · 8 tasks, corrupted L2avg accuracy: 75.44 -> 78.78Improved
Add semantic divergence term Div within ADS (ablation)BD-Merging · 8 tasks, corrupted L2avg accuracy: 76.28 -> 78.78Improved
Add opinion conflict term Conf within ADS (ablation)BD-Merging · 8 tasks, corrupted L2avg accuracy: 76.95 -> 78.78Improved
Add prediction sharpness term Sharp within ADS (ablation)BD-Merging · 8 tasks, cleanavg accuracy: 86.68 -> 87.15Improved
Add inverse correlation loss L_Inv (ablation)BD-Merging · 8 tasks, corrupted L2avg accuracy: 77.43 -> 78.78Improved
Use BD-Merging instead of Twin-Merging on unseen tasksCLIP ViT-B/32 merged model · 4 unseen tasksavg accuracy: 53.03 -> 55.01Improved
Merge models (BD-Merging) vs. pretrained CLIP on unseen tasksCLIP ViT-B/32 · 4 unseen tasksavg accuracy: 56.99 -> 55.01Hurt
Increase corruption severity from L1 to L3BD-Merging (layer-wise) · 8 tasksavg accuracy: 83.31 -> 75.36Hurt
Use BD-Merging layer-wise instead of AdaMerging w/ Surgery under corruption L2CLIP ViT-B/32 · 8 tasks, corrupted L2avg accuracy: 74.33 -> 78.78Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy